During the last decade, the rapid growth of remote sensing applications has been a strong catalyst to research about the relationships between landscapes, biodiversity, and ecosystem functioning. Despite some important advances regarding the role of spatial feedbacks in structuring terrestrial ecosystems, such observations are lagging in aquatic systems because of the inherent difficulty of assessing their landscape composition and complexity. To investigate the technical capacity of high-resolution remote sensing to fill this gap, we used simulated landscapes to emulate the main sources of remote sensing noise affecting aquatic scenes at multiple sensor resolution and vegetation organisation levels. Multiple informational complexity metrics were then computed over these landscapes to study their behavior across gradients of organisation, resolution, and noise, assessing the limits of their usage in a monitoring context. To ensure the simulated landscapes were representative of natural gradients, simulated noise was also added to real high-resolution remote sensed landscapes to compare the complexity response curves. Preliminary results suggests that some metrics evaluated on very high-resolution images (<5m resolution) are less robust to environmental noise than on high or medium resolution images, while some complexity metrics such as those related to shape tend to perform well across all noise treatments. These findings represent an important step towards studying landscapes complexity changes across time and space and how it affects functioning and resilience.

Primary Presenter: de Grandpré Arthur, Université du Québec à Trois-Rivières (arthur.de.grandpre@uqtr.ca)

Authors:

Arthur de Grandpré, Université du Québec à Trois-Rivières  (arthur.de.grandpre@uqtr.ca)

Christophe Kinnard, Université du Québec à Trois-Rivières  (christophe.kinnard@uqtr.ca)

Andrea Bertolo, Université du Québec à Trois-Rivières  (andrea.bertolo@uqtr.ca)